Welcome to monkeys

What monkeys helps you do

monkeys is a place to run task-focused AI agents. Describe the result you need, choose an agent, and follow its work in a conversation. Start on the home page or browse the Agent Hub when you want to compare available agents.

Agents and models

An agent combines instructions, capabilities, and a way of working for a particular kind of task. A model is the underlying AI engine. The task composer normally asks you to choose an agent or a Best category route. When an agent allows model choice, a separate model control appears; fixed-model and existing self-hosted agents may not expose one.

• Choose a named agent when you want a specific published setup.

• Choose Best Chat or another available Best category when you want monkeys to route to an eligible agent.

• Check the agent page and composer controls for the capabilities available in that task.

From prompt to result

A task starts with your prompt and any context you add. While the agent runs, the conversation can show progress, questions, messages, and files. A finished message and a delivered file are different outputs: read the response for conclusions, and open or download file cards when the agent produces a reusable artifact.

1. Describe the outcome, audience, and constraints.

2. Add only the files the agent needs.

3. Review questions or progress instead of sending duplicate requests.

4. Check the final response and any separate deliverables.

Keep work organized

Signed-in tasks appear in the Tasks history. You can open another conversation while a run continues and return to see its live or final state. Use the same conversation for follow-up work; start a new task when you want a clean context or a different agent.

Next steps

Follow the quickstart, then read Models and agents and Files and deliverables. For spending and plan behavior, see Credits before starting a long task.